Dynamic Evolution of Bimetallic Single-Atom Catalysts during Carbon Dioxide Reduction Probed by Operando X‑ray Spectroscopy

Abstract

Electrochemical CO₂ reduction (CO₂RR) offers a promising strategy to convert CO₂ into high-value chemicals and fuels. However, probing the mechanism of dynamic catalyst reconstruction remains a longstanding challenge, which calls for operando time-resolved and quantitative spectroscopic methods. Here, we employed operando high-energy-resolution fluorescence-detected X-ray absorption spectroscopy (HERFD-XAS) to monitor real-time structural evolution of bimetallic FeCu single-atom catalysts (SACs) on nitrogen-doped carbon (FeCu–N–C) as a model single-atom catalyst for the CO₂RR. FeCu–N–C exhibits a 2-fold increase in Faradaic efficiency for CO formation and an order-of-magnitude enhancement in the CO partial current density at the same applied potential, when compared to monometallic Fe SAC (Fe–N–C). To elucidate the origin of this performance enhancement, operando HERFD-XAS was used to directly probe the dynamic evolution of metal oxidation states and coordination environments under operating conditions, revealing a dynamic reconstruction of isolated Fe and Cu single atoms into catalytically active metallic clusters. In contrast, the Fe SAC control remains atomically dispersed over the course of CO₂RR. The atomic-scale structural insights are further corroborated by aberration-corrected scanning transmission electron microscopy. These findings shed light on the synergistic role of Cu in modulating the dynamic reconstruction of single-atom Fe sites and provide fundamental insights for the rational design of high-performance single-atom catalysts for CO₂ electroreduction.

Introduction

The electrochemical reduction of CO₂ into high-value chemicals offers a promising approach for addressing pressing climate change issues. − Achieving efficient CO₂RR remains challenging due to low selectivity, high overpotentials required to activate the thermally stable CO₂ molecule, and competition with hydrogen evolution (HER) in aqueous media. Consequently, there is a pressing need for catalysts that combine high activity, selectivity, and stability. While Au and Ag catalysts are highly effective for CO production at low overpotentials, their high cost limits practical applications, highlighting the importance of developing inexpensive, efficient, and earth-abundant alternatives.

Recently, there have been substantial advancements in the development of transition-metal-based catalysts. Among these, SACs have emerged as a particularly promising class of materials, as they maximize catalytic efficiency and approach complete metal utilization. , In addition to high efficiency, SACs offer additional benefits, including tunable catalyst supports and atomic control over the active sites. SACs also often exhibit catalytic behavior that differs markedly from their nanoparticle or bulk counterparts due to their isolated active sites and distinctive electronic structures. These unique characteristics offer enhanced activity, selectivity, and tunability, positioning SACs as powerful platforms for next-generation catalytic applications. SACs have been extensively explored for the CO₂RR, yielding a broad variety of products, including high-value chemicals such as syngas (CO/H₂), ethanol and ethylene, underscoring their potential industrial implementation. Among them, Fe based SACs exhibit promising activity and selectivity for the conversion of CO₂ to CO, whereas Cu SACs are reported to promote CO₂ conversion toward multicarbon products but at a high overpotential. , Thus, it is possible to design bimetallic SACs to further improve catalyst activity and stability, relative to their monometallic counterparts.

Metal–organic frameworks (MOFs) are a highly versatile class of crystalline porous materials, distinguished by their exceptionally high surface areas and periodically ordered pore architectures. A defining advantage of MOFs lies in the tunability of both their metal nodes and their organic linkers, which allows precise control over the local coordination environment and enables the creation of abundant, well-defined metal-binding sites. These characteristics render MOFs particularly attractive for the stabilization of isolated metal species and have led to their widespread use as precursors for SAC design.

Upon thermal conversion, the long-range order of MOFs can be translated into heteroatom-doped carbon frameworks while partially preserving the intrinsic porosity of the parent structure. This transformation substantially broadens the applicability of MOF-derived SACs, especially in electrochemical systems where efficient charge transport, structural robustness, and operational durability are critical. Pyrolysis routes enable the production of porous carbons with high specific surface areas. − Moreover, the interconnected pore networks inherent to MOFs and their carbon derivatives facilitate rapid mass transport, enhancing the adsorption, diffusion, and activation of CO₂, other reactants, and key intermediates, thereby maximizing the accessibility and utilization of atomically dispersed active sites.

In this work, we employed the synthetic strategy using a sacrificial Zn-based MOF mixed with transition metal salts. This approach enables the uniform incorporation of metal species as atomically dispersed sites within the MOF-derived, N-doped carbon matrix, while effectively suppressing the formation of metallic nanoparticles. Moreover, operando synchrotron HERFD-XAS was employed to discern the distinct structural and electronic behaviors of bimetallic versus monometallic single-atom catalysts. A comprehensive analysis and evaluation of the X-ray near-edge structure spectra (XANES) and extended X-ray absorption fine structure (EXAFS) spectra, complemented by detailed electron microscopy, provided direct insights into the nature of the catalytically active sites and allowed us to establish clear correlations between the metal center identity and the observed catalytic performance.

Notably, HERFD-XAS, which probes a narrow emission band (e.g., Cu K_α1) and provides a substantially enhanced energy resolution (∼0.5 eV) relative to conventional XAS (∼1.5 eV), allowed for the sensitive detection of pre-edge features, which are otherwise spectroscopically undetectable. Leveraging this capability, we reveal that the presence of Cu induces the reconstruction of bimetallic FeCu–N–C from isolated single-atom sites into clustered species under CO₂RR conditions, whereas monometallic Fe–N–C retain their atomically dispersed structure during operation. These findings provide valuable insights for the rational design and optimization of SACs for CO₂RR.

Results and Discussion

Leveraging the structure ZIF-8, we synthesized Fe, Cu, and FeCu SACs, with precise atomic dispersion of metal centers within a nitrogen-doped porous carbon matrix. The catalysts are denoted as FeCu–N–C, Fe–N–C, Cu–N–C, and undoped-N–C corresponding to FeCu bimetallic SACs, monometallic Fe, Cu SACs, and undoped SACs, respectively. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images were acquired with a fifth-order aberration-corrected STEM with sub-Ångström spatial resolution (Figure a). Low-magnification HAADF-STEM images reveal the uniform morphology of the carbon catalyst supports at the nanoscale (Figure a), while energy-dispersive X-ray (STEM–EDX) mapping (Figures b–d and S1) confirms the uniform distribution of Fe and Cu throughout the material. Residual Zn is also detected, consistent with previous studies and related to incomplete Zn volatilization during carbonization. Atomic-scale HAADF-STEM (Figure e) further identifies isolated metal atoms, providing direct evidence for the successful formation of SACs.

1.

(a) HAADF-STEM image of FeCu–N–C, with EDX elemental mapping of (b) FeCu (composite colors), (c) Cu (red) and (d) Fe (green). (e) Atomic-scale resolution HAADF-STEM image of FeCu–N–C with representative isolated metal atoms highlighted by red circles and corresponding EDX spectrum (inset). (f) Powder XRD of FeCu–N–C, Fe–N–C, and Cu–N–C. High-resolution XPS N 1s spectra of: (g) Cu–N–C, (h) Fe–N–C and (i) FeCu–N–C systems.

To gain further structural insights, multiple analytical techniques were used. Surface composition and chemical state of the catalysts were examined by X-ray photoelectron spectroscopy (XPS). The overall XPS spectra confirmed the presence of C, N, O and the corresponding metal dopants (Fe or/and Cu), verifying the successful incorporation of both nitrogen and metals within the carbon matrix. The relative atomic fractions of C, N, and O were comparable across all samples, with representative values of 79–84% for C, 10–14% for N, and 3–4.5% for O (Table S1, Supporting Information), while the total metal content was approximately ∼0.1 at. % for each catalyst. Consistent with previous reports, the high-resolution N 1s spectra were deconvolved into four components at 398.4, 399.8, 400.9, and 402.9 eV, assigned to pyridinic, metal-coordinated, pyrrolic, and oxidized N species, respectively. Both monometallic and bimetallic SACs exhibited similar proportions of metal–N coordination environments, with metal-coordinated nitrogen accounting for approximately 7.5–10% (Figure g–i). Moreover, the C 1s spectra featured a dominant peak at 284.6 eV corresponding to C–C/CC bonds, along with higher-binding-energy contributions at 285.5, 286.5, 287.8, and 289.5 eV attributed to C–N, C–O, CO, and O–CO species, respectively (Figure S2). ,

N₂ adsorption–desorption isotherm measurements were conducted to characterize the textural properties of the catalysts, including their specific surface area, pore volume, and pore size distribution. Although FeCu–N–C exhibits a slightly higher t-plot surface area, the overall surface area and pore volume differences among the samples are relatively small (Figures S9–S12, Table S2). The crystallinity of the metal-doped ZIF-8 structures was evaluated by X-ray diffraction (XRD) analysis. As shown in Figure S3, the XRD patterns of the mono and bimetallic ZIF-8 samples (Fe, Cu, and FeCu) closely match those of pristine ZIF-8, indicating that the structural framework remains intact. These results not only confirm the successful synthesis of ZIF-8 but also demonstrate that metal incorporation does not compromise the phase purity of the ZIF-8 lattice. Furthermore, the XRD peaks observed at 2θ values of 7.5°, 10.5°, 12.8°, 14.7°, 16.4°, 18.1°, 22.2°, and 24.6° correspond to the (011), (002), (112), (022), (013), (222), (114), and (233) crystallographic planes of the ZIF-8 topology, respectively. After pyrolysis of the metal-containing ZIF-8, the XRD patterns revealed features typical of amorphous carbon, showing broad amorphous peaks near 25.5° and 43.4°, corresponding to the (002) and (100) graphite planes; no reflections associated with crystalline phases were detected (Figure f). To further verify the origin of the diffraction features, iron oxide was physically mixed with undoped N–C as a control. While the resulting composite showed clear diffraction peaks attributable to iron oxide, no such peaks were detected in the pure undoped N–C sample (Figure S13). Although the absence of diffraction peaks suggests the absence of crystalline metal or metal oxide nanoparticles, XRD analysis alone cannot unequivocally confirm the formation of SACs, particularly in the case of highly dispersed or ultrasmall species. Therefore, the identification of SACs was further supported by complementary HAADF-STEM analysis (Figure a), which provided direct evidence for the atomic dispersion of the metal species. Additional morphological characterization for all catalysts was employed by SEM (Figure S4), which revealed that the catalysts predominantly consist of rhombic dodecahedral structures with an average size of approximately ∼85 nm with no observable granules, indicating atomic-level metal dispersion and preservation of the original ZIF-8 framework.

Since both Fe and Cu are well-established active centers for CO₂ reduction, an H-cell configuration was used to electrochemically verify the presence of atomically dispersed metal sites and subsequently evaluate the catalytic performance of FeCu–N–C, along with the monometallic controls Fe–N–C and Cu–N–C. In addition, previous theoretical studies on FeCu–N–C have shown, through Bader charge analysis, that electron transfer from the carbon substrate to the Fe centers can modulate the electronic structure of Fe sites and promote CO₂ reduction activity. It was also observed that the d-band values for bimetallic single-atom catalysts are less negative than for the monometallic counterparts, and this can favor a higher activity. It is possible that the bimetallic clusters possess a d-band value even more favorable toward the CO₂RR, explaining the increased activity in the present work. All potentials reported herein were converted to the reversible hydrogen electrode (RHE) scale and corrected for iR drop unless otherwise specified (see Experimental Section for details). As shown in Figure a, FeCu–N–C delivers a high CO Faradaic efficiency (FE_CO) of 82.6% at −0.6 V, accompanied by a current density of −3.5 mA cm^–2. Notably, the FE_CO remains above 65% over a wide potential range from −0.6 to −0.8 V, while current densities exceed −5 mA cm^–2 across this potential window. In contrast, monometallic Fe–N–C catalyst exhibits a substantially lower FE_CO of approximately 45% and a modest current density of ∼−0.6 mA cm^–2 at −0.6 V (Figure b). The incomplete FE balance may originate from the relatively low current density, where the low product generation rate increases the uncertainty associated with product quantification; monometallic Cu–N–C SAC exhibit an even lower FE_CO at −0.6 V, which may attribute to low Cu sites in Cu–N–C (Figure S5). Cu is not intrinsically selective toward CO production because it tends to retain *CO on the surface instead of releasing it as gaseous CO, while the HER pathway competes strongly for active sites and electrons. ,

2.

(a,b) Gas product distributions for (a) FeCu–N–C and (b) Fe–N–C during CO₂ electroreduction (left Y-axis: FE of CO %; right Y-axis: total current density). (c) Partial CO current densities of FeCu–N–C and Fe–N–C as a function of applied potential, in CO₂-saturated 0.1 M KHCO₃. (d) Stability tests of FeCu–N–C and pure Fe–N–C were conducted at −0.6 V for over 10 h (left Y-axis: FE of CO %; right Y-axis: Current density black dashed line: FeCu–N–C, blue solid line: Fe–N–C).

Figure c summarizes the CO partial current density, a key metric for evaluating intrinsic CO₂-to-CO catalytic activity. FeCu–N–C exhibits a markedly higher CO partial current density, approximately an order of magnitude greater than that of Fe–N–C, clearly underscoring its superior intrinsic activity. Stability tests (Figure d) further demonstrate that FeCu–N–C sustains both reasonable FE_CO and current density for at least 10 h without discernible degradation, whereas Fe–N–C undergoes rapid deactivation within the first hour.

Taken together, the potential-dependent performance, CO partial current density and long-term stability consistently demonstrate that Cu incorporation substantially enhances the CO₂RR performance of Fe–N–C. Notably, the FeCu–N–C enables low-overpotential operation (−0.6 V), high CO production rates and stable activity over extended reaction (10 h), demonstrating promise for practical CO₂RR applications. These observations naturally prompt a key mechanistic question: how does Cu promote the activity and stability of Fe–N–C? Addressing this issue necessitated an in-depth mechanistic investigation using operando HERFD-XAS (vide infra).

Motivated by promising structural characteristics and catalytic performance, we employed operando HERFD-XAS to gain deeper insight into the dynamic structural evolution of the catalyst and to elucidate how Cu modulates Fe–N–C activity under operating CO₂RR conditions. Operando HERFD XANES of Cu K-edge of FeCu–N–C were compared with those of Cu reference standards (Cu foil, Cu₂O, and CuO in Figure a). Upon stepping the applied potential from the open circuit potential (OCP, around 0.6 V) to −1.0 V, the Cu K-edge spectra exhibit a progressive increase in pre-edge intensity, accompanied by a pronounced decrease in white-line intensity, gradually approaching the spectral features of metallic Cu (red arrow near 8980 eV, Figure a). The gradual attenuation of the feature near 8995 eV further confirms the reduction of Cu­(II) to Cu(0). , These results indicate a clear potential-dependent transformation of Cu species from isolated Cu­(II) sites to metallic Cu, which is generally the catalytically active state for CO₂ reduction.

3.

Operando HERFD-XANES spectra of FeCu–N–C examining (a) Cu K-edge and (b) Fe K-edge across different reducing potentials (the dashed vertical lines mark the range of the LCA analysis). LCA analysis of the relative fraction of (c) metallic Cu and (d) metallic Fe for FeCu–N–C.

Complementary operando HERFD XANES measurements of the Fe K-edge reveal that Fe­(III) species are progressively reduced to Fe­(II) as the applied potential becomes more negative, as evidenced by the comparison with reference spectra (Figure b).

To quantitatively track the evolution of metal valence state under CO₂RR conditions, linear combination analysis (LCA) was performed on operando XANES spectra. Cu foil, Cu₂O, and FeCu–N–C were selected as reference compounds representing Cu(0), Cu­(I), and Cu­(II) species, respectively. Similarly, Fe foil, FeO, and FeCu–N–C were used as reference compounds for Fe(0), Fe­(II), and Fe­(III) species, respectively. For the Cu edge (Figure c), LCA indicates that nearly 70% of atomically dispersed Cu is reduced to metallic Cu at −0.5 V, increasing to ∼80% at −1.0 V, demonstrating a monotonic potential-driven conversion from Cu­(II) to Cu(0). In parallel, LCA of the Fe K-edge (Figure d) reveals that at −0.3 V, ∼5% of single-atom Fe is reduced to metallic Fe and ∼50% to Fe²⁺, with the metallic fraction increasing to ∼20% at −0.5 V. Notably, at −0.3 V, where the catalyst contains ∼35% metallic Cu but only ∼5% metallic Fe, no measurable CO₂RR activity is observed. In contrast, at −0.5 V, where the catalyst comprises ∼80% of metallic Cu and ∼20% metallic Fe, clear CO₂RR activity trends emerge.

To understand the reversibility of this structural reconstruction, LCA was also performed on samples exposed to air following CO₂RR (electrolyte removal and subsequent air exposure). The results show that metallic Cu species reoxidize, while most Fe species similarly revert to the oxidized state (Figure c,d), indicating a possibly reversible reconstruction process under ambient conditions.

While XANES and LCA provide insights into oxidation state evolution and quantitative trends in single atoms to metallic transformations, distinguishing isolated high-valent single-atom sites from oxidized metallic clusters remains challenging due to their similar oxidation states and overlapping XANES signatures. Therefore, EXAFS analysis is required to probe the local coordination environments and unambiguously identify the nature of the metal species, particularly following air exposure.

EXAFS analysis provides direct insight into the local coordination environment of the absorbing atoms and enables identification of the active structural motifs under CO₂RR conditions. To better visualize these structural changes, we employed wavelet transform EXAFS (wt-EXAFS), which simultaneously resolves information in both k-space and R-space.

As shown in Figure b, wt-EXAFS reveals that FeCu–N–C develops distinct scattering features at 2.3 Å in R space at −0.5 V, consistent with the formation of Fe–Cu and/or Fe–Fe in the presence of Cu. On the other hand, Fe–N–C (Figure e) maintains the characteristic single-atom features under identical electrochemical conditions, with no evidence of metal clustering.

4.

Wavelet transform analysis of operando Fe K-edge EXAFS for (a–c) FeCu–N–C and (d–f) Fe–N–C acquired at −0.5V vs. RHE and the OCP during CO₂ electroreduction and after post-CO₂RR air exposure.

To examine the reversibility of this structural reconstruction, wt-EXAFS were further collected for FeCu–N–C after air exposure. The dominant scattering feature is observed at ∼1.6 Å in R space, which is substantially shorter than the Fe–Fe distance in metallic Fe, indicating that metallic clustering does not persist upon air exposure. To exclude the possibility that the air-exposed species correspond to Fe₂O₃ rather than isolated Fe sites, the spectra were compared with a Fe₂O₃ reference (Figure S6). The Fe₂O₃ reference exhibits a pronounced peak at ∼2.9 Å in R-space and prominent intensity in the 7–10 Å^–1 region in k-space, features that are notably absent in the air-exposed FeCu–N–C sample.

The absence of Fe₂O₃ signatures, together with the re-emergence of coordination features characteristic of isolated Fe sites, suggests that FeCu–N–C reverts to an atomically dispersed configuration after air exposure. However, due to the limited sensitivity of EXAFS, this observation alone is insufficient to definitively prove the reversibility of the structural evolution. To provide more comprehensive evidence, atomic-resolution STEM characterization was performed on FeCu–N–C after the CO₂RR. As shown in Figure S7b, the majority of the catalyst consists of carbon nanoparticles with sizes ranging from 150 to 250 nm, while isolated single-atom sites are still clearly observed at atomic-scale resolution (Figure S7a). In addition, reconstructed nanoclusters with sizes of approximately 100 nm are also detected (Figure S8a). EDX elemental mapping reveals that these nanoclusters contain both Fe and Cu, with Fe being the dominant component (Figure S8b–d).

We found both single-atom sites and clusters after reaction, although the majority of species shift from clusters during CO₂RR to SAC after air exposure, it only indicates that the structural reconstruction induced during CO₂RR is partially reversible. While we can confirm that reconstruction toward Fe–Cu cluster formation occurs under the CO₂RR, some of the catalysts revert to atomically dispersed Fe sites after air exposure, underscoring the dynamic nature of the Fe–Cu active sites. Taken together, these correlations strongly suggest that the reconstructed bimetallic Fe–Cu clusters evolved from FeCu–N–C constitute the catalytically active sites during CO₂RR.

Conclusions

In conclusion, we demonstrate that bimetallic FeCu–N–C, synthesized using a ZIF-8 precursor, undergoes a pronounced potential-induced reconstruction into clustered active sites during CO₂ reduction, whereas similarly synthesized monometallic Fe–N–C retain their single-atom configuration across the investigated potential range. Atomic-resolution STEM imaging confirms the presence of atomically dispersed metal sites in the as-prepared catalysts. By leveraging the unique capabilities of operando HERFD-XAS, we directly monitored, in real time, the evolution of metal oxidation states and local coordination environments in FeCu–N–C under CO₂RR conditions. Our results reveal that Cu plays a crucial role in inducing the reconstruction of Fe–N–C into Fe–Cu clusters during operation, and that these clusters can partially revert to single-atom sites upon air exposure after the CO₂RR. As a model bimetallic single-atom catalyst system, FeCu–N–C provides fundamental insights into atomic-scale reconstruction phenomena under electrochemical conditions. This work establishes design principles for stabilizing or deliberately exploiting dynamic structural reconstruction in SACs, thereby enabling their practical deployment in CO₂RR. More broadly, it highlights how atomic-level dynamics can be harnessed to reduce catalyst cost, while simultaneously enhancing activity and stability.

Experimental Section

Materials

2-Methylimidazole (C₄H₆N₂, >98%) and sodium benzoate (C₇H₅NaO₂, >98%) were purchased from Tokyo Chemical Industry Co., Ltd. (TCI). Methanol (CH₃OH, >99.8%), zinc nitrate hexahydrate (Zn­(NO₃)₂·6H₂O, 99%), ferrous sulfate heptahydrate (FeSO₄·7H₂O, >99%), copper­(II) sulfate pentahydrate (CuSO₄·5H₂O, >98%) were purchased from Sigma-Aldrich. All the chemicals were analytical grade and used without further purification.

Synthesis of FeCu–N–C and Fe–N–C

Zn­(NO₃)₂·6H₂O (0.01 mol) and sodium benzoate (1.5 mmol, C₇H₅NaO₂) was dissolved in 100 mL of methanol under stirring for 20 min to form a clear solution. In a separate beaker, the metal precursors (0.4 mmol of FeSO₄·7H₂O and 0.4 mmol CuSO₄·5H₂O) together with 2-methylimidazole (0.056 mol) were dissolved in 400 mL of methanol under simultaneous stirring and ultrasonication for 30 min. The two solutions were then rapidly combined, and the resulting mixture was stirred at room temperature for 24 h. The precipitated solid was collected by centrifugation at 10,000 rpm for 5 min, washed with methanol and dried overnight (referred to as FeCu/ZIF8). Similar procedures were followed for the synthesis of monometallic ZIF8 precursors (Fe/ZIF8 and Cu/ZIF8, respectively).

The metal-containing ZIF-8 precursors (including pristine ZIF-8) was heated in a tubular furnace at 900 °C for 3 h under a continuous nitrogen flow, with a heating rate of 3 °C min^–1. After pyrolysis, the resulting porous carbon was mechanically ground using an agate mortar. The samples are denoted as FeCu–N–C (SACs/PC), Fe–N–C (SACs/PC), Cu–N–C (SACs/PC) and undoped-N–C (SACs/PC).

Materials Characterization

X-ray diffractograms (XRD) were obtained using a D8 Advance diffractometer (Bruker AXS GmbH) with a Cu Kα radiation (λ = 1.54 Å). The powder samples were deposited on background-free Si wafers and scanned between 10 and 85° 2θ, in steps of 0.02°, and a rate of 0.3 s per step. X-ray photoelectron spectroscopy (XPS) was conducted using a Thermo Scientific Nexsa G2 spectrometer at an operating pressure of approximately 1 × 10^–9 Torr. A monochromatic Al Kα X-ray (1486.6 eV) source was used, and the photoelectrons were collected from a 400 μm diameter analysis spot at a 90° emission angle and a 54.7° source-to-analyzer angle. Pass energies of 200 eV for survey scans and 50 eV for high-resolution scans were used. A flood gun was used to neutralize charging in nonconductive samples. Deconvolution of the spectra was performed with the CasaXPS software. Field emission scanning electron microscopy (FE-SEM) images were obtained with a Zeiss-Gemini-500-FESEM at a low accelerating voltage of 3 kV to reduce the excitation volume and enhance resolution.

A Thermo-Fisher FEI Spectra 300 (Kraken), equipped with an ultrahigh-brightness cold field emission gun and spherical aberration corrector, was used to collect all high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images, as well as energy dispersive X-ray spectroscopy (STEM–EDX) elemental maps. The samples were dispersed in analytic-grade MeOH and dropcast onto Au TEM grids and dried at 75 °C for 12 h prior to electron microscopy analysis. The images were collected using the following parameters: 300 kV accelerating voltage; 30 mrad convergence angle, with 60 mrad inner and 200 mrad outer collection angles; 70 μm C2 aperture; 94 mm camera length; 61 pm STEM resolution. The Fe Kα (green) and Cu Lα (red) energy lines were used for EDS mapping.

Electrochemical CO₂ Reduction (CO₂RR)

A custom-fabricated polyether ether ketone (PEEK) H-cell was used to evaluate the electrochemical CO₂ reduction reaction (CO₂RR) performance. A Selemion ion-exchange membrane (Anion AMV-N, Selemion, NARA Cell Tech Corp. South Korea) was used to separate the working and counter electrode chambers. Carbon paper (Toray Paper, TGP-H-120, 5% wet-proofing) served as the substrate for the working electrode, with a geometric active area of 0.49 cm², onto which catalyst ink was spray-coated at a loading of 6 mg cm^–2. The catalyst ink was prepared by dispersing 3.0 mg of catalyst powder (FeCu–N–C or Fe–N–C) and 1 μL of Nafion ionomer (D520CS.5 wt %, Ion Power Inc. USA) in 1 mL of isopropanol (IPA), followed by sonication for 1 h to obtain a homogeneous suspension. A platinum mesh and an Ag/AgCl reference electrode (WPI, Dri-ref, 2 mm diameter, 3 M KCl) were used as the counter and reference electrodes, respectively. A potentiostat (Biologic SP-200) was used to control the potential. During CO₂RR measurements, CO₂ gas was continuously purged into the catholyte at a flow rate of 20 sccm. The electrolyte was prepared by bubbling CO₂ through 15 mL of 0.1 M KHCO₃ until saturation was achieved. The bicarbonate buffer effectively maintained the solution pH, and no appreciable pH change was observed during CO₂ saturation. Electrochemical measurements were conducted after 10 min of electrolysis at each applied potential, at which point the current had reached a steady state. All applied potentials were converted to the reversible hydrogen electrode (RHE) scale using the following equation (eq )­

$$E(\mathrm{v}\mathrm{s}\mathrm{R}\mathrm{H}\mathrm{E},iR\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d})=E\left(\mathrm{v}\mathrm{s}\frac{\mathrm{A}\mathrm{g}}{\mathrm{A}\mathrm{g}\mathrm{C}\mathrm{l}}\right)+0.210+0.059\times \mathrm{p}\mathrm{H}-i{R}_{\mathrm{s}}$$ 1

where the current (i) is defined as negative under reduction conditions. The gaseous products generated during electrolysis (H₂ and CO) were detected and quantified by gas chromatography (GC, SRI Instruments, MG#5). H₂ was analyzed using a thermal conductivity detector (TCD), while CO was detected using a flame ionization detector (FID). The Faradaic efficiency (FE) of each product was calculated according to the following equation (eq )­

$$\mathrm{F}\mathrm{E}(\%)=\frac{i_{\mathrm{p}\mathrm{a}\mathrm{r}\mathrm{t}\mathrm{i}\mathrm{a}\mathrm{l}}}{i_{\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}}\times 100=\frac{V_{\mathrm{p}\mathrm{a}\mathrm{r}\mathrm{t}\mathrm{i}\mathrm{a}\mathrm{l}}\times Q\times \frac{nF}{RT}}{i_{\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}}\times 100$$ 2

where V _partial is volume concentration, Q is the gas flow rate, n is the number of electrons transferred per mole of product formed, F is the Faraday constant, R is the ideal gas constant, and T is the temperature.

Operando HERFD-XAS Measurements

X-ray absorption spectroscopy (XAS) measurements were performed at the PIPOXS beamline of the Cornell High Energy Synchrotron Source (CHESS). Fe K-edge and Cu K-edge XAS spectra were collected in high-energy-resolution fluorescence-detected (HERFD) mode using a cryogenically cooled Si(311) double-crystal monochromator. A customized thin-layer liquid electrochemical cell, previously reported by our group, was employed for operando electrochemical CO₂ reduction reaction (CO₂RR) measurements. Operando HERFD-XAS data were acquired under chronoamperometric (CA) control in CO₂-saturated (sat.) 0.1 M KHCO₃ electrolyte, electrochemical measurements were performed using a potentiostat (Biologic SP-200). Samples labeled as “air-exposed” were obtained by carefully removing all electrolyte from the electrochemical cell without disturbing the sample position. The acquired XAS spectra were processed and analyzed using the Athena software package. Wavelet-transformed EXAFS (wt-EXAFS) analysis was carried out using HAMA Fortran software with a Morlet wavelet applied to k ²-weighted χ­(k) spectra.

N₂ Adsorption Isotherm Measurements

Nitrogen adsorption–desorption isotherms were acquired using a Micromeritics ASAP 2460 surface area and porosity analyzer with ultrahigh-purity N₂ gas (99.999%) as the adsorbate. Prior to analysis, the carbon samples were degassed under vacuum (∼10^–4 mbar) at 140 °C for 24 h to remove any physisorbed species. To minimize the dead volume within the sample tubes, glass filler rods were inserted, while the free space was automatically determined before each measurement using high-purity helium gas (99.999%). The adsorption measurements were conducted at 77 K using liquid nitrogen over a relative pressure (P/P ₀) range from 10^–5 to 0.99, with an equilibration time of 30 s per point. Pore size distributions were calculated from the N₂ adsorption isotherms at 77 K using density functional theory (DFT) with a carbon-based slit-pore model and employed in the adsorption analysis software provided by the manufacturer (Micromeritics). More details regarding the DFT analysis can be found in previous studies by Thommes and Cychosz and Vedaprakash et al.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.chemmater.6c00765.

• Figure S1: STEM–EDS spectrum confirms the presence of all constituent elements in the FeCu–N–C sample. The presence of Na is adventitious; Figure S2: the presence of Au is the result of using Au TEM grids. Deconvoluted high-resolution C 1s spectra of (a) FeCu–N–C (b) Cu–N–C and (c) Fe–N–C; Figure S3: XRD patterns of Fe/ZIF8, Cu/ZIF8, FeCu/ZIF8 and ZIF8; Figure S4: FE-SEM images of (a) Fe–N–C, (b) Cu–N–C and (c) FeCu–N–C; Figure S5: gas product distributions for (a) Cu–N–C; (b) undoped-N–C during CO₂ electroreduction (left Y-axis: FE of CO %; right Y-axis: total current density); Figure S6: wavelet transform analysis of operando Fe K-edge EXAFS for Fe₂O₃ reference; Figure S7: HAADF-STEM image of FeCu–N–C after CO₂RR (a) atomic scale resolution (b) low magnification (with representative isolated metal atoms highlighted by red circles); Figure S8: (a) HAADF-STEM image of FeCu clusters found after CO₂RR with a week fully air exposure, with EDX elemental mapping of (b) Cu (red), (c) Fe (green) and (d) FeCu (composite color); Figure S9: N₂ porosimetry measurements for: (a) Fe–N–C, (b) Cu–N–C and (c) FeCu–N–C, respectively; Figure S10: density functional theory (DFT) pore size distribution of (a) Fe–N–C, (b) Cu–N–C and (c) FeCu–N–C based on the N₂–carbon, slit-cylinder kernel; Figure S11: N₂ porosimetry measurements for: (a) Fe/ZIF8, (b) Cu/ZIF8 and (c) FeCu/ZIF8, respectively; Figure S12: density functional theory (DFT) pore size distribution of (a) Fe/ZIF8, (b) Cu/ZIF8 and (c) FeCu/ZIF8 based on the N₂–carbon, slit-cylinder kernel; Figure S13: XRD patterns for undoped-N–C and undoped-N–C mixed with iron oxide; Table S1: atomic composition of all carbon catalysts based on XPS surveys; Table S2: surface area analysis parameters of all carbon catalysts (PDF)

∥.

J.J., N.C., and R.S. contributed equally.

The authors declare no competing financial interest.

Published as part of Chemistry of Materials special issue “Bifunctional Electrocatalysts”.